Sense My FEUP - Mobility forecasting based on GPS location tracking

PART I: EDA

Package installation.

Connecting the database.

General EDA.

Getting data.

df_segments_length <- data.frame()
df_segments_length_tm <- dbGetQuery(con, "SELECT session_id, SUM(length) as distance, travelmode
                                      FROM danielasocas.segments_travelmode as segments 
                                      WHERE segments.travelmode = 'car' OR segments.travelmode = 'bus'
                                      GROUP BY session_id, travelmode")

df_segments_length <- dbGetQuery(con, "SELECT session_id, SUM(length) as distance
                                      FROM danielasocas.segments_travelmode as segments 
                                      WHERE segments.travelmode = 'car' OR segments.travelmode = 'bus'
                                      GROUP BY session_id")

df_segments <- subset(df_segments_length, distance < IQR(distance,0.75) )

ggplot(df_segments, aes(distance)) +
  geom_histogram(bins = 10)

ggplot(df_segments_length, aes(travelmode, distance, color = travelmode)) +
  geom_boxplot()

ggplot(df_segments, aes(distance, color = travelmode)) +
  geom_histogram(binwidth = 1)

ggplot(df_segments, aes(travelmode, distance, color = travelmode)) +
  geom_boxplot()

df_segments_timestamp <- data.frame()
df_segments_timestamp <- dbGetQuery(con, "SELECT session_id, time_start,time_end, length, travelmode
                                      FROM danielasocas.segments_travelmode as segments
                                      WHERE segments.travelmode = 'car' OR segments.travelmode = 'bus'")

Getting data per edge.

POINTS per edge.

Getting the points filtered by only segments with travelmode = car or bus.

SESSION per edge.

Getting the sessions filtered by only segments with travelmode = car or bus.

df_sessions_edge_osm_cars <- data.frame()
df_sessions_edge_osm_cars <- dbGetQuery(con, "SELECT way_id, COUNT(DISTINCT session_id) AS sessions
                                   FROM sensemyfeup_raw.osmlocation as osm 
                                          INNER JOIN danielasocas.segments_travelmode as segments 
                                            USING (session_id)
                                          INNER JOIN sensemyfeup_raw.session_offsets 
                                            USING (session_id)
                                       WHERE (osm.seconds + session_offsets.clock_offset_seconds BETWEEN segments.seconds_start AND seconds_end) AND
                                            segments.travelmode = 'car' OR segments.travelmode =  'bus'
                                       GROUP BY way_id
                                       ORDER BY sessions DESC")

save(df_sessions_edge_osm_cars, file="df_sessions_edge_osm_cars.Rda")

Plots edges and loading.

POINTS

p1 <- subset(df_osm_edge, points < mean(points)) 
summary()

ggplot(p1, aes(points)) + geom_histogram(binwidth = 10) +
  xlim(0,1000)

ECDF

SESSIONS

p1 <- subset(df_osm_edge, sessions < mean(sessions)) 
summary()

ggplot(p1, aes(sessions)) + geom_histogram(binwidth=1) +
  scale_x_continuous(breaks = seq(0,8,1))

ECDF

list_osm_edge <- df_osm_edge[, 1]

fun.ecdf <- ecdf(df_osm_edge$sessions)

my.ecdf <- fun.ecdf(sort(df_osm_edge$sessions))

my_ecdf_df <- data.frame(my.ecdf)

my_ecdf_df$sessions <- sort(df_osm_edge$sessions)

tail(my_ecdf_df, 400)

P <- ecdf(df_osm_edge$sessions)

plot(P, log="x", xlim=c(1, max(df_osm_edge$sessions)))

Map points/sessions per edge.

Mapping

# Showing Map.
m 

Writing the results

Disconnecting Databases.